Multiple Interpretations and Constraints of Causative Serial Verb Constructions in Korean

eScholarship provides open access, scholarly publishingservices to the University of California and delivers a dynamicresearch platform to scholars worldwide.

Proceedings of the Annual Meeting of theBerkeley Linguistics Society

UC Berkeley

Peer Reviewed

Title:Multiple Interpretations and Constraints of Causative Serial Verb Constructions in Korean

Journal Issue:Proceedings of the Annual Meeting of the Berkeley Linguistics Society, 38

Author:Lee, Juwon

Publication Date:2014

Permalink:http://escholarship.org/uc/item/03x1x85t

Keywords:Berkeley Linguistics Society

Copyright Information:All rights reserved unless otherwise indicated. Contact the author or original publisher for anynecessary permissions. eScholarship is not the copyright owner for deposited works. Learn moreat http://www.escholarship.org/help_copyright.html#reuse

http://escholarship.org




http://escholarship.org/uc/bling_proceedings

http://escholarship.org/uc/bling_proceedings

http://escholarship.org/uc/ucb

http://escholarship.org/uc/bling_proceedings?volume=38;issue=38

http://escholarship.org/uc/search?creator=Lee%2C%20Juwon

http://escholarship.org/uc/item/03x1x85t

http://www.escholarship.org/help_copyright.html#reuse

Multiple Interpretations and Constraints of Causative Serial Verb

Constructions in Korean

Author(s): Juwon Lee

Proceedings of the 38th Annual Meeting of the Berkeley Linguistics

Society (2014), pp. 288-306

General Session and Thematic Session on Language Contact

Editors: Kayla Carpenter, Oana David, Florian Lionnet, Christine Sheil,

Tammy Stark, Vivian Wauters

Please contact BLS regarding any further use of this work. BLS retains

copyright for both print and screen forms of the publication. BLS may be

contacted via http://linguistics.berkeley.edu/bls/.

The Annual Proceedings of the Berkeley Linguistics Society is published online

via eLanguage, the Linguistic Society of America's digital publishing platform.

http://linguistics.berkeley.edu/bls/

http://www.elanguage.net/

Multiple Interpretations and Constraints of Causative Serial Verb Constructions in Korean JUWON LEE The University of Texas at Austin 1 Introduction In this paper I discuss the light verb ha ‘do’ in Korean, which I show forms a range of uses in various constructions (see the basic properties of light verbs in Butt and Geuder (2001), Butt (2004) and Korean light verbs in Choi and Wechsler (2001), Lee (2011), inter alia). In particular, I aim to elucidate what is involved in the multiple interpretations of the causative serial verb construction (SVC) like (1). (1) Mary-ka khephi-lul ttukep-key hay mek-ess-ta. Mary-Nom coffee-Acc hot-Key do.Comp eat-Pst-Dec ‘Mary made the coffeej hot, and then ate itj/ tried to eat itj.’ ‘Mary brewed the coffeej hot, and then ate itj/ tried to eat itj.’ Most research on causative constructions is only concerned with the canonical resultative reading (see Comrie 1981, Song 1996, among others). However, Korean causative constructions are in fact very ambiguous as illustrated in the English translations for (1). The light verb hay in (1) can have the meaning of made or brewed and the final verb mek-ess-ta can have the resultative reading ate or the purposive reading tried to eat, in whose event structure an unspecified causing subevent of eating necessarily occurs but the result of eating is not entailed, but just intended.

Addressing the origins and constraints of the multiple interpretations of causative SVCs, I first explore the three kinds of constructions headed by the light verb ha. Examples of the constructions are illustrated in (2).

(2) a. Jane-i khephi-lul hay-ss-ta. Jane-Nom coffee-Acc do-Pst-Dec ‘Jane brewed/ drank the coffee.’

288

Juwon Lee

b. Jane-i khephi-lul ttukep-key hay-ss-ta. Jane-Nom coffee-Acc hot-Key do-Pst-Dec ‘Jane made/ brewed the coffee hot.’ c. Mary-ka khephi-lul hay mek-ess-ta. Mary-Nom coffee-Acc do.Comp eat-Pst-Dec ‘Mary brewed the coffeej, and then ate itj.’ In (2a), the related issue is how the specific meaning (brew or drink) of the light verb ha gets picked up in the construction (developed from Lee (2011, 2012)). In (2b), the light verb is not just giving the standard small clause type reading for the causative, but also picking up on the predicates related to the common noun. Many scholars assume that in periphrastic causatives we have a structure like [[sc NP XP] CAUSE] where the NP is not directly predicated of by the causal verb, and thus this should never happen. But in the small clause structure, the XP “transmits” the associated predicate (e.g. brew) of the NP (e.g. khephi ‘coffee’) up to CAUSE. Or, it could be that CAUSE takes directly two arguments (the NP and XP). The latter analysis appears to be more perspicuous. In the SVC (2c), an important question is why the light verb hay is restricted to a certain associated predicate (brew, but not drink), and how other related constraints of SVC should be reflected in explicit semantic analysis of hay. These issues are interconnected with each other in the causative SVCs like (1). Note that the sentences in (2) can also have their relevant purposive interpretations, which will be discussed in the sections that follow.

Based on the properties of the constructions, I then suggest a formal analysis of the constructions in the framework of Minimal Recursion Semantics (MRS) (Copestake et al. 2006) of Head-driven Phrase Structure Grammar (HPSG) (Pollard and Sag 1994, Sag et al. 2003). 2 Background: Purposive and Resultative Readings In this section I present a puzzle about Korean resultatives, namely the fact that the putative result states need not actually obtain. The Korean equivalent of the following English sentence is acceptable: He wiped the table clean, but the table is not clean. The crucial question that naturally arises is what exactly the meaning of the resultative construction is, especially when the result does not necessarily obtain. I will argue that in fact these constructions are systematically ambiguous between two readings: a canonical resultative reading in which the result does obtain and an additional reading in which the result is simply intended, with a semantics much like a purposive construction.

First, in the typical purposive construction (3a), the putative result of napcakha-key ‘flat-Key’ is cancellable. In (3b), it shows that the construction is compatible with the modifier intentionally, but not with accidently.

289

Multiple Interpretations and Constraints of Causative Serial Verb Constructions in Korean

(3) a. John-i soy-lul napcakha-key mantul-kiwihay twutulki-ess-ta. John-Nom metal-Acc flat-Key make-to hammer-Pst-Dec kulena soy-ka napcakha-ci anh-ta. but metal-Nom flat-Comp Neg-Dec ‘John hammered the metalj to make itj flat, but the metal is not flat.’ b. John-i soy-lul napcakha-key mantul-kiwihay John-Nom metal-Acc flat-Key make-to ilpwule/ #wuyenhi twutulki-ess-ta. intentionally/ accidently hammer-Pst-Dec ‘John intentionally/ #accidently hammered the metalj to make itj flat.’ The event cancellation in (3a) indicates that the result state of the Adj-key phrase is not entailed in the purposive construction. In (3b), the contrast in modification shows that purposive constructions require some intentionality on the part of the subject of the matrix clause.

The typical Korean resultative construction with the secondary predicate (napcakha-key ‘flat-Key’) in (4a) (see more Korean resultatives in Wechsler and Noh (2001)) also allows the result cancellation like the purposive in (3a). However, in (4b), the modification of accidently is permitted unlike (3b). (4) a. John-i soy-lul napcakha-key twutulki-ess-ta. John-Nom metal-Acc flat-Key hammer-Pst-Dec kulena soy-ka napcakha-ci anh-ta. but metal-Nom flat-Comp Neg-Dec (lit.) ‘John hammered the metal flat, but the metal is not flat.’ = ‘John hammered the metalj to make itj flat, but itj is not flat.’ b. John-i soy-lul napcakha-key ilpwule/ wuyenhi John-Nom metal-Acc flat-Key intentionally/ accidently twutulki-ess-ta. hammer-Pst-Dec ‘John intentionally/ accidently hammered the metal flat.’ The cancellation in (4a) suggests that the construction has the purposive reading; since accidently is not compatible with a purposive reading, the construction in (4b) has the normal resultative reading. In these two interpretations, the main verb meaning corresponds to the causing subevent of the construction’s event structure and it must be entailed in the construction with a secondary predicate.

The two crucial differences between purposive and resultative readings in terms of event cancellation and adverb modification can be illustrated as in (5). (5) Purposive

reading Resultative reading

Cancellation of relevant result state Yes No Modification by ‘accidently’ No Yes

290

Juwon Lee

If a resultative reading manifests itself in the context of the modification of

wuyenhi ‘accidently’, then we can predict that the construction with the adverb cannot allow the result cancellation. This is confirmed in the following: (6) John-i soy-lul napcakha-key wuyenhi twutulki-ess-ta. John-Nom metal-Acc flat-Key accidently hammer-Pst-Dec #kulena soy-ka napcakha-ci anh-ta. but metal-Nom flat-Comp Neg-Dec ‘John accidently hammered the metal flat, #but the metal is not flat.’

In sum, the Korean constructions that have generally been considered as resultatives are actually ambiguous between purposive and resultative readings. This is a common property of structures that can entail a result state in Korean, and in some constructions one or the other reading of the structure is ruled out. With this background, I now turn to Korean light verb constructions (LVCs). 3 Qualia Light Verb ha ‘do’ In this section, I show that the light verb ha in an LVC can have a specific meaning according to its common noun object. Then based on the properties of purposive and resultative readings discussed in the previous section, I argue that if the light verb ha is interpreted as having the meaning of a change-of-state verb, the change-of-state meaning of ha also has purposive and resultative readings.

In the following minimal pair in (7), we see that the light verb ha receives its specific meaning depending on its common noun object (cf. Im and Lee 2004 and see Lee 2011, 2012). (7) Yenghi-ka khephi-lul/ #mwul-ul hay-ss-ta. Yenghi-Nom coffee-Acc/ water-Acc do-Pst-Dec ‘Yenghi brewed/ drank the coffee.’ In (7), when the object is khephi ‘coffee’, ha is interpreted as having the meaning of brew or drink. However, the unacceptability with mwul ‘water’ shows that it is not that every common noun can appear as the object of the construction.

If the light verb can have the meanings of associated predicates, we can predict that the different interpretations of ha should behave differently regarding aspect. This is verified in the following sentences with maney ‘in’: (8) Yenghi-ka pap-ul/ khemphuthe-lul han sikan maney hay-ss-ta. Yenghi-Nom rice-Acc/ computer-Acc one hour in do-Pst-Dec ‘Yenghi cooked the rice in an hour.’ (telic or ingressive reading) ‘Yenghi used the computer in an hour.’ (ingressive reading)

291


Thus certain common nouns have information about their related predicates. Pustejovsky (1991) refers to this relation as cospecification; just like a verb can select for its argument type (e.g. kick selects its argument like ball, but not happiness), an argument also can select its particular associated predicates (e.g. ball may select its predicate like kick, but not read). The associated predicate information is then included in the Qualia Structure of a lexical item (Pustejovsky 1991). In the Qualia Structure, the Telic Role has values about purpose and function of object (e.g. read for novel), and the Agentive Role has values on factors involved in the origin of an object (e.g. write for novel).

Building on qualia structure, I suggest that Korean common nouns have the dual semantic components, the first of which is the meaning of the common noun itself (e.g. rice relation for pap ‘rice’) and the second of which is the associated predicate meanings (e.g. cook relation for pap ‘rice’) (see Copestake and Briscoe (1995) for qualia roles in feature structure). However, the common nouns like mwul ‘water’ have no value for their qualia roles. Although the predicate masi- ‘drink’ appears to be a good candidate for the telic role of mwul ‘water’, there seems to be no grammatical evidence to verify if masi- ‘drink’ is really used in grammar as a telic role for mwul ‘water’; masi- ‘drink’ seems to not yet be conventionalized as an associated predicate of mwul ‘water’.

When the light verb ha is interpreted as having an associated predicate (e.g. brew), it can have the relevant purposive reading as in (9a) in addition to the usual resultative reading as in (9b). (9) a. Yenghi-ka achim-ey khephi-lul hay-ss-ta. Yenghi-Nom morning-in coffee-Acc do-Pst-Dec kulena khephi-ka mantul-e ci-ci anh-ass-ta. but coffee-Nom make-Comp Pass-Comp Neg-Pst-Dec ‘Yenghi tried to brew a coffee, but a coffee was not made.’ b. Yenghi-ka achim-ey wuyenhi khephi-lul hay-ss-ta. Yenghi-Nom morning-in accidently coffee-Acc do-Pst-Dec #kulena khephi-ka mantul-e ci-ci anh-ass-ta. but coffee-Nom make-Comp Pass-Comp Neg-Pst-Dec ‘Yenghi accidently brewed a coffee, #but a coffee was not made.’ In syntax, an adverb can appear in between the common noun object and the light verb as shown in (8), which indicates that the verb phrases of the qualia-ha constructions should be analyzed syntactically rather than lexically. 4 Causative Light Verb ha ‘do’ In this section, I discuss a use of ha ‘do’ as marking causative constructions like the use of make in English causative constructions, and the ambiguity of the secondary predicate in the causative constructions.

292

Juwon Lee

Causative constructions are normally classified into two types: lexical causative and periphrastic (syntactic) causative (see Comrie 1981, Song 1996, among others). For instance, the Korean lexical causative in (10) (where the causative dependent morpheme -i is attached to the verb stem) describes an event of direct causation: i.e. the subject is necessarily the agent who dried the clothes by e.g. operating a drying machine or hanging the clothes on a drying rack. In contrast, the periphrastic causative with the result XP in (10) does not entail a direct causation, although the interpretation of a direction causation is possible; Tom can make someone else dry the clothes (i.e. an indirect causation). In this paper, I focus on the direct causation reading of the Korean periphrastic causative construction. (10) Tom-i caki os-ul mal-i-ess-ta/ malu-key hay-ss-ta. Tom-Nom self clothes-Acc dry-Caus-Pst-Dec/ dry-Key do-Pst-Dec ‘Tomi dried hisi clothes.’/ (lit.) ‘Tomi did hisi clothes dry.’ = ‘Tomi made hisi clothes dry.’

If the periphrastic causative sentence in (10) can be interpreted as its purposive reading, we can predict that the construction should allow the relevant result state cancellation. Also if it is interpreted as its resultative reading with the modification of accidently, it is predicted that the cancellation is not allowed. These two predictions are borne out in (11a) and (11b), respectively. (11) a. Tom-i caki os-ul malu-key hay-ss-ta. Tom-Nom self clothes-Acc dry-Key do-Pst-Dec kulena os-i malu-ci anh-ass-ta. but clothes-Nom dry-Comp Neg-Pst-Dec ‘Tomi tried to make hisi clothes dry, but they were not dry.’ b. Tom-i wuyenhi caki os-ul malu-key hay-ss-ta. Tom-Nom accidently self clothes-Acc dry-Key do-Pst-Dec #kulena os-i malu-ci anh-ass-ta. but clothes-Nom dry-Comp Neg-Pst-Dec ‘Tomi accidently made hisi clothes dry, #but they were not dry.’

One interesting similarity of the light verb ha in the periphrastic causative construction to the light verb ha in the qualia-ha construction is that they get their meaning from their complement. As already shown in section 3, the specific meaning of ha of qualia-ha construction is determined by the common noun object (e.g. cook from rice object). The meaning of ha of periphrastic causative construction (i.e. use as make) is determined by the existence of a secondary predicate (e.g. malu-key ‘dry-Key’); while the exact meaning of ha cannot be identified unless it can be inferred from the wider context in which the construction appears, ha here corresponds to the unspecified causing subevent in the event structure of the construction. In the event structure, the unspecified

293


causing subevent necessarily occurs; what contributes to the ambiguity of the construction is the ambiguity of the secondary predicate between realized result state (in resultative reading) and intended result state (in purposive reading).

5 Qualia-Causative Light verb ha ‘do’ The Korean light verb ha is very ambiguous, as shown above. It can be either qualia light verb, as discussed in section 3, or causative light verb, as shown in section 4, each of which then has its purposive and resultative readings. In this section, I discuss a sort of mixed use of the light verb (namely, qualia-causative light verb ha) in a single construction, and then its theoretical implication for the syntactic analysis of the construction in question.

In (12), the light verb ha ‘do’ can be the normal causative light verb or the qualia-causative light verb (the quasi-depictive reading with the telic role drink ‘Jane drank the coffee hot’ is not discussed in this paper). (12) Jane-i khephi-lul ttukep-key hay-ss-ta. Jane-Nom coffee-Acc hot-Key do-Pst-Dec. ‘Jane made the coffee hot/ tried to make the coffee hot.’ ‘Jane brewed a coffee hot/ tried to brew a coffee hot.’ In the normal causative reading of (12), the only relevant result is that the coffee becomes hot. However, in the qualia-causative reading, two results are involved (i.e. the creation of a coffee and creation of the property of being hot). Here the qualia-causative light verb ha gets its meaning from both the common noun object and the XP; the light verb ha corresponds to the combination of the event of brewing a coffee and the unspecified causing subevent of making the object hot. In the event structure of the construction, if the associated results are all realized, then the sentence has the resultative reading, but if some result is not realized, but only intended, the sentence has the purposive reading (roughly corresponding to ‘Jane tried to brew a coffee hot’) with the following three possible situations: i) a cold coffee was made, ii) a hot tea was made, and iii) a cold tea was made. In any case, the result of the construction (i.e. a hot coffee) is not realized.

According to small clause analysis of periphrastic causative construction, the object and the secondary predicate are syntactically grouped together to form a small clause (i.e. a predication), and then the causative verb combines with the small clause. This analysis seems to have no problem for the normal causative reading of the sentence (12).

However, for the sentence (12) with the qualia-causative interpretation, the small clause analysis appears to be not perspicuous relatively (although it is not impossible) since the XP (i.e. the head of the small clause) should “transmit” the qualia meaning from the common noun object (i.e. the complement of XP) up to the light verb ha when the light verb ha combines with the small clause in syntax. Rather than this transmission mechanism of the small clause analysis, it is more

294

Juwon Lee

perspicuous for the light verb ha to combine directly with the NP and XP in syntax, and thus gets its qualia meaning directly from the NP. For the sake of a theoretical consistency, it is also better to analyze the construction (12) with the normal causative interpretation in much the same way.

In the next section, the causative construction (12) is then combined with an SVC (resulting in causative SVC). This combination creates an interesting restriction on the possible interpretations of the light verb in the causative SVC. 6 Causative SVC The core concept of an SVC is to serialize the events of component verbs of the construction and thus to conceptualize the component events as a single, unified event, as exemplified in (13a) (see more e.g. in Collins (1997), Aikhenvald (2006), Kim (2010)). SVCs are generally under the iconicity constraint: i.e. the subevent of the first verb (V1) must occur before the subevent of the second verb (V2). So, the sequential order of the component verbs and their corresponding subevents are basically parallel. The iconicity constraint leads (13b) to be ungrammatical. (13) a. kunye-ka kheik-ul cip-e mek-ess-ta. she-Nom cake-Acc pick.up-Comp eat-Pst-Dec ‘She picked up the cakej, and then ate itj/ tried to eat itj.’ b. *kunye-ka kheik-ul mek-e cip-ess-ta. she-Nom cake-Acc eat-Comp pick.up-Pst-Dec (int.) ‘She picked up the cakej, and then ate itj/ tried to eat itj.’ The meaning of picking up a cake which is already eaten is implausible. The world knowledge is, however, based on the assumption that the SVC in (13b) is iconic-constrained. If the iconicity constraint is not relevant (so it is possible for the first verb to represent the subevent that happens after the subevent of the second verb), the SVC should be fine, and must have the intended reading; but it cannot. Thus the iconicity constraint is an underlying property of SVCs; by contrast coordinations are not under the iconicity constraint.

Another important fact of SVC is that although the usual change-of-state verbs in Korean do not necessarily entail their relevant result states (or result objects), Korean SVCs do not allow the cancellation of the result state(s) related to the first verb event (more generally, non-final verb events); but the event of the final verb of an SVC can be cancelled just like the normal change-of-state verbs.

The following two constructions in (14) have the two fundamental properties of SVC (iconicity constraint and no cancellation of result state(s) associated with V1), which strongly indicate that they are really a type of SVC in Korean. The result state of the secondary predicate related to V1 is not cancellable, either. Moreover, only the agentive role brew is appropriate for the meaning of the V1 light verb hay suggesting that a light V1 in an SVC can only take on the agentive reading (not telic reading) for its object (see Lee 2011, 2012):

295


(14) a. Mary-ka khephi-lul hay mek-ess-ta. Mary-Nom coffee-Acc do.Comp eat-Pst-Dec ‘Mary brewed the coffeej, and then ate itj/ tried to eat itj.’ b. Mary-ka khephi-lul ttukep-key hay mek-ess-ta. Mary-Nom coffee-Acc hot-Key do.Comp eat-Pst-Dec ‘Mary made/brewed the coffeej hot, and then ate itj/ tried to eat itj.’ The SVC (14b) is the causative SVC which is the combination of the causative construction in (12) and the SVC in (14a) (cf. Aikhenvald, 2006: 16). In the next section, I formalize the semantic analysis of the multiple interpretations of the causative SVC. 7 A formal analysis First, resultative and purposive meanings (i.e. CONTENTs) of the change-of-state verb tha- ‘brew’ can be declared like the following: (15) a. Resultative CONTENT of ‘brew’:

LBL h1 LBL h4RELS < ARG0 e1 , ARG0 e2 ,

ARG1 h2 ARG1 ARG2 h3 ARG2

LBL h5 LBL h7,ARG0 e3 ARG0 e4ARG1 h6 AR

resultative_cause_rel _action_of_brewing_rel

ij

_become_rel _brewed_rel

>

G1 _ _ _ _ _ _

HCONS < HARG h2 , HARG h3 , HARG h6LARG h4 LARG h5 LARG h7

jqeq rel qeq rel qeq rel

>

b. Purposive CONTENT of ‘brew’:

LBL h1 LBL h4RELS < ARG0 e1 , ARG0 e2 ,


LBL h5 LBL h7,ARG0 e3 ARG0 e4ARG1 h6 ARG

purposive_cause_rel _action_of_brewing_rel

ij

_purpose_rel _become_rel

LBL h9, >ARG0 e51 h8 ARG1

_ _ _ _ _ _ _ _HCONS < HARG h2 , HARG h3 , HARG h6 , HARG h8

LARG h4 LARG h5 LARG h7 LARG h9

_brewed_rel

jqeq rel qeq rel qeq rel qeq rel

>

The purposive CONTENT (15b) is specified as having [_purpose_rel] in its REL(ATION)S.

296

Juwon Lee

As for common nouns, it is claimed that they include the QUALIA-ST(RUCTURE) in CONT(ENT), which in turn has the AGENTIVE and TELIC attributes, and the QUALIA list is also posited whose value is the sum of the values of the AGENTIVE and TELIC attributes (see Lee 2011, 2012). For instance, the common noun khephi ‘coffee’ can have the following feature structure: (16) khephi ‘coffee’:

[ ] [ ][ ]

PHON < >

LTOP h1HOOK < >INDEX

RELS < LBL h1 >ARG0

CONT HCONS < >AGENTIVE RELS < , ,...>

QUALIA-ST TELIC RELS < ,

cnkhephi

j_coffee_rel

j

A _cause_rel _action_of_brewing_rel

B _cause_rel _a

< > < [ ],...>

QUALIA

ction_of_drinking_rel

A B

> ⊕

In (16), khephi ‘coffee’ has one agentive role (i.e. [RELS <[_cause_rel], [_action_of_brewing_rel]…>]), and one telic role (i.e. [RELS <[_cause_rel], [_action_of_drinking_rel],…>]). These two qualia roles are underspecified with respect to resultative or purposive CONTENT; thus for example either of (15) can be the value of the AGENTIVE attribute.

The core meaning of the light verb ha in a qualia-ha construction comes from a qualia role of the common noun object as represented in (17a). The qualia light verb ha in (17a) should have the corresponding V1 form in (17b), which is used in an SVC.

(17) a. The qualia ha-1 in non-SVC:

PHON < 1 >ARG-ST < NP , NP QUALIA-ST QUALIA <..., 1 ,...>

CONTENT 1

tr-light-vha

i j

−

>

b. The qualia hay-1 in SVC:

[ ]

PHON < 1 >FORM SUBJ < NP >

QUALIA-ST AGENTIVE < 1 >COMPS < NP >

CONTENT 1

haye

iresultative

j

− −

297


In (17a), a value of the QUALIA of the object NPj is structure-shared with the value of the CONTENT of the light verb ha. In other words, the light verb requires an object that has at least one value for QUALIA. This requirement prevents common nouns like mwul ‘water’ from appearing as an object of the light verb. Since the non-final light verb in SVC should have a resultative agentive role, hay-1 in (17b) requires a common noun object that has a resultative value for the AGENTIVE of the QUALIA-ST.

Also secondary predicates can have purposive or resultative reading, which is reflected in the two feature structures below:

(18) a. ttukep-key-1 ‘hot-Key’:

PHON < 1 >LTOP h1

HOOK < INDEX e1 >XARG

_ _ _ _LBL h1 LBL h3CONT RELS < , >ARG0 e1 ARG0 e2ARG1 h2 ARG1

_ _HCONS < HARG0 h2 >

LARG1 h3

ttukep key

j

become rel hot rel

j

qeq rel

− −

b. ttukep-key-2 ‘hot-Key’:

PHON < 2 >LTOP h1

HOOK < INDEX e1 >XARG

_ _ _ _ _ _LBL h1 LBL h3 LBL h5CONT RELS < , , >ARG0 e1 ARG0 e2 ARG0 e3ARG1 h2 ARG1 h4 ARG1

HCON

ttukep key

j

purpose rel become rel hot rel

j

− −

_ _ _ _S < HARG0 h2 , HARG0 h4 >

LARG1 h3 LARG1 h5

qeq rel qeq rel

A lexical rule can be posited to license (18b) from (18a), which can also be derived by another lexical rule taking ttukep ‘hot’ as the input.

The heavy verbs used as V2 (generally, the final verb) of an SVC should have different lexical items from those used in non-SVCs:

298

Juwon Lee

(19) a. lexeme mek-1 ‘eat’ in non-SVC:

[ ]PHON < 1 > ARG-ST < NP , NP >

LTOP h1HOOK INDEX e1

CONT LBL h1 LBL h4RELS < ARG0 e1 , ARG0 e2 ,... >


mekacci j

_cause_rel _action_of_eating_rel

ij

−

b. lexeme mek-2 ‘eat’ as V2 in SVC:

[ ]

PHON < 2 >

HEAD [FORM ]ARG-ST < 1NP , 2 NP , V SUBJ < 1NP > >

COMPS < 2 NP ,...

LTOP h1HOOK INDEX e1

CONT LBL h1RELS < ARG0 e1 ,

ARG1 h2ARG2 h3

mek

eacci j i

j

_cause_rel _action_of_eat

− − >

LBL h4ARG0 e2 ,... >ARG1 ARG2

ing_rel

ij

Both the verbs in (19) can have resultative or purposive meaning, as represented with the underspecified feature [RELS<[_cause_rel], [_action_of_eating_rel],…>]. A main difference is that the accusative object of the V2 in (19b) is shared by the V1 which is in the ARG(UMENT)-ST(RUCTURE) of the V2.

The following phrase structure rule for the combination of component verbs of SVC is proposed (see Lee 2011, 2012). In (20), the V1 is one of the complements of the V2, and accusative object is shared by the V1 and V2. This shared object and the possible non-shared complements (i.e. A and B) are passed up to the resulting combination, where the constructional meaning (C-CONT) is added. Now the VP (i.e. [NP [hay V2]]) of (14a) can be analyzed like (21).

299


(20) Head-Obj-Share-SVC:

SUBJ < 2 >COMPS < 4 >

HOOK 3CONT RELS

HCONS

LTOP h4 HOOK 3 INDEX e3_ _ _LBL h4

C-CONT RELS < ARG0 e3 >ARG1 h5ARG2 h6

HCONS < HARG h5L

hd obj share svc

A B

G C DH E F

lead to rel

G

qeqH

− − −

⊕ ⊕

⊕ ⊕

⊕ ⊕

FORM

SUBJ < 2 >COMPS < 4 [ ]

1

HOOK LTOCONT

, HARG h6 > ARG h1 LARG h2

nonstative ve

acc A

qeq

−

− >⊕ →

[ ][ ]

SUBJ < 2 >COMPS 4 1

,

HOOK LTOP h2P h1 CONT RELS

RELS HCONS HCONS

nonstative v

B

DC F

E

−

< > ⊕ ⊕ < >

H

(21) [[khephi-lul]NP [hay mek-ess-ta] Head-Obj-Share-SVC]hd-comp-ph

PHON < > SUBJ < 1 >COMPS < >DTRS < 2 , 6

PHON < > SUBJ < 1 >COMPS < 2 >DTRS < 3 , 5

PHON < > SUBJ < 1

HD-DTR 6

HD-DTR 5

hd comp phkhephi lul hay mek ess ta

hd obj share svchay mek ess ta

nonstative vmek ess ta

− −− − −

>

− − −− −

>

−− −

NP >PHON < 1 > FORM SUBJ < 1NP >

RELS < LBL h1 >COMPS < 2 NP , 3 ARG0 COMPS < 2 NP >

QUALIA-ST AGENTIVE 4RELS

CONT 4

ihaye

i_coffee_rel

Cjj

j resultativeA

− −

< >

[ ]

[ ]

>

LBL h5CONT RELS < , ARG0 e2 ,... >

ARG1 ARG2

CONT RELS _ _ _

_action_of_eating_rel

B _cause_relij

lead to rel A B

< > ⊕ ⊕ [ ]CONT RELS _ _ _lead to rel A B C

< > ⊕ ⊕ ⊕

300

Juwon Lee

The final CONTENT of the VP in (21) means that the subject’s action of brewing a coffee caused the creation of a coffee (i.e. the subject brewed a coffee), and then ate the coffee (i.e. resultative reading) or tried to eat the coffee (i.e. purposive reading).

Now, in addition to the normal causative light verb ha-2 in (22a) for the VP (i.e. [NP [XP [ha]]]) of the non-SVC causative construction, the causative V1 light verb hay-2 in (22b) is required for the VP (i.e. [NP [XP [hay V2]]]) of the causative SVC having the normal causative interpretation. (22) a. The causative ha-2 in non-SVC:

PHON < 2 >FORM

ARG-ST < NP , NP , XP LTOP h1CONT HOOK XARG

LTOP h2 HOOK INDEX e1

LBL h2 CONT RELS < ARG0 e1 ,

ARG1 h3ARG2 h4

tr-light-vha

key

i j j

_cause_rel _causal_even

−−

>

LBL h5 ARG0 e2 > ARG1 ARG2

HCONS < LARG h3 , LARG h4 >HARG h5 HARG h1

tuality_rel

ij

_qeq_rel _qeq_rel

b. The causative hay-2 in SVC:

PHON < 2 >FORM SUBJ < NP >

FORM

COMPS < NP , XP CONT LTOP h1HOOK XARG

LTOP h2 HOOK INDEX e1

LBL h2 RELS < ARG0 e1CONT

ARG1 h3ARG2 h4

haye

i

keyresultative

jj

_cause_rel

−−

− >

LBL h5 , ARG0 e2 >

ARG1 ARG2


_causal_eventuality_rel

ij

_qeq_rel _qeq_rel

301


Also in (23a), the qualia-causative light verb for [NP [XP [ha]]] is presented, and in (23b) the V1 qualia-causative light verb for [NP [XP [hay V2]]] is posited.

(23) a. The qualia-causative ha-3:

PHON < 3 >

LBL h1RELS ARG0 e1 ,...

ARG-ST < NP ,NP AGENTIVE <..., ,...> ,ARG1 h2 ARG2 h3

HCONS FORM -

XP CONT HOOK LTO

tr light vha

_cause_rel

Ai j

Bkey

− −−

< >

[ ]

[ ]

P h5

HOOK LTOP h1

LBL h4 CONT RELS < >ARG0 e2ARG1 ARG2

HCONS < LARG h2 , LARG h3 > HARG h4 HARG h5


A

ij

_qeq_rel _qeq_relB

>

⊕ ⊕

b. The qualia-causative hay-3:

PHON < 3 >

LBL h1ARG-ST < NP ,NP AGENTIVE <..., RELS ARG0 e1 ,... ,...> ,

ARG1 h2 ARG2 h3

HCONS FORM -

XP CONT

hayresultative

_cause_rel

Ai j

Bkey

re

− < >

[ ]

[ ]

HOOK LTOP h5

HOOK LTOP h1

LBL h4 CONT RELS < >ARG0 e2ARG1 ARG2


sultative


A

ij

_qeq_rel _qeq_relB

>

⊕

⊕

302

Juwon Lee

In (23a), each of the agentive role and the XP can have either purposive or resultative meaning, but in (23b), the agentive role and the XP are restricted to resultative meanings. Equipped with the lexical items and the phrase rule (20), we can analyze the VP (i.e. [NP [XP [hay V2]]]) of the causative SVC with the qualia-causative reading, as in the following: (24) [[khephi-lul]NP [[ttukep-key] [hay mek-ess-ta]Head-Obj-Share-SVC] hd-comp-ph]

PHON < >SUBJ < 1 >COMPS < >DTRS < 2 , 7

PHON < >SUBJ < 1 >COMPS < 2 >DTRS < 4, 6

PHON <

HD-DTR7

HD-DTR6

hd comp phkhephi lul ttukep key hay mek ess ta

hd comp phttukep key hay mek ess ta

hd obj share svchay mek

− −− − − −

>− −

− − −

>− − −

−

[ ]

[ ]

>SUBJ < 1 >COMPS < 2, 4 >DTRS < 3, 5

PHON < >SUBJ < 1NP >

PHON < >

HD-DTR5 RELS < _ _ >COMPS < 2NP

AGENTIVE RELS ,...

ess ta

nonstative vmek ess ta

i

khephi lul

E coffee relj resultative

A _cause_rel

−

>−

− −

− < > < >

PHON < 3 >SUBJ < 1NP >

PHON < >, 3 >COMPS < 2NP , 4 >CONT

RELS

CONT RELS < >_

CONT RELS <

hay

ittukep key

resultativej

D_causal_A B eventuality rel

C _cause

− − ⊕

[ ] [ ]

[ ]

[ ]

, ,...>

CONT RELS _ _ _

CONT RELS _ _ _

_rel _action_of_eating_rel

lead to rel A B C

lead to rel A B C D

< >⊕ ⊕ ⊕

< >⊕ ⊕ ⊕ ⊕

[ ]CONT RELS _ _ _lead to rel A B C D E

< >⊕ ⊕ ⊕ ⊕ ⊕

The final CONTENT in (24) means that the subject’s action of brewing a coffee and making it hot caused the creation of a hot coffee (i.e. the subject brewed the coffee hot), and then the subject’s action of eating the hot coffee caused the result that the coffee became eaten (i.e. the resultative reading of the V2) or the result of eating the coffee is simply intended (i.e. the purposive reading of the V2). 8 Conclusion The specific meaning of the qualia light verb ha in qualia-ha construction comes from a qualia role (e.g. brew or drink) of common noun object. In the causative construction, the qualia-causative light verb ha has the mixed meaning of a qualia role and the causative light verb (whose meaning is influenced by the XP); the common noun object of the causative construction can be directly predicated of by the causal verb. In the context of SVC, the meaning of the qualia light verb hay is restricted to the resultative meaning of an agentive role. Finally, in the

303


causative SVC, the qualia-causative light verb hay has the mixed meaning of the causative light verb and the resultative interpretation of an agentive role. The semantic analysis provided here may be applied to other complex predicates in Korean and other languages (e.g. Tamil, Japanese, and Chinese) that allow some kinds of event cancellation. References Aikhenvald, Alexandra Y. 2006. Serial Verb Constructions in Typological

Perspective. In Aikhenvald and Dixon, eds., Serial Verb Constructions: A Cross-Linguistic Typology, 1-68. Oxford: Oxford University Press.

Butt, Miriam. 2004. The Light Verb Jungle. Harvard Working Papers in

Linguistics 9:1-49. Butt, Miriam, and Wilhelm Geuder. 2001. On the (Semi)Lexical Status of Light

Verbs. In Norbert Corver and Henk van Riemsdijk, eds., Semilexical Categories: On the content of function words and the function of content words, 323-370. Berlin: Mouton de Gruyter.

Choi, Incheol and Stephen Wechsler. 2001. Mixed Categories and Argument

Transfer in the Korean Light Verb Construction. In Proceedings of the 8th International Conference on Head-driven Phrase Structure Grammar, 103-120.

Collins, Chris. 1997. Argument Sharing in Serial Verb Constructions. Linguistic

Inquiry 28(3):461-497. Comrie, Bernard. 1981. Language Universals and Linguistic Typology: Syntax

and Morphology. Chicago: University of Chicago Press. Copestake, Ann, Dan Flickinger, Carl Pollard, and Ivan A. Sag. 2006. Minimal

Recursion Semantics: An Introduction. Research on Language and Computation 3:281-332.

Copestake, Ann and Ted Briscoe. 1995. Semi-productive Polysemy and Sense

Extension. Journal of Semantics 12(1): 15-67. Im, Seohyun and Chungmin Lee. 2004. Combination of the Verb ha- ‘do’ and

Entity Type Nouns in Korean: A Generative Lexicon Approach. [in Korean] Language and Information 8(1):77-100.

Kim, Jong-Bok. 2010. Argument Composition in Korean Serial Verb

Constructions. Studies in Modern Grammar 61:1-24.

304

Juwon Lee

Lee, Juwon. 2011. Two Types of Korean Light Verb Constructions in a Typed Feature Structure Grammar. In Proceedings of the Workshop on Multiword Expressions: from Parsing and Generation to the Real World, 40-48.

Lee, Juwon. 2012. Change of State Verb and Syntax of Serial Verb Constructions

in Korean: An HPSG Account. Coyote Papers: Working Papers in Linguistics 20:57-65.

Pollard, Carl and Ivan A. Sag. 1994. Head-Driven Phrase Structure Grammar.

Stanford, CA: CSLI. Pustejovsky, James. 1991. The Generative Lexicon. Computational Linguistics

17(4):409-441. Sag, Ivan A., Thomas Wasow, and Emily M. Bender. 2003. Syntactic Theory: A

Formal Introduction, 2nd edition. Stanford, CA: CSLI. Song, Jae Jung. 1996. Causatives and causation: A universal-typological

perspective. London and New York: Addison Wesley Longman. Wechsler, Stephen and Bokyung Noh. 2001. On resultative predicates and clauses:

Parallels between Korean and English. Language Sciences 23:391-423. Juwon Lee The University of Texas at Austin Department of Linguistics College of Liberal Arts Building 4.304 Austin, TX 78712 [email protected]

305

306

Documents

Multiple Interpretations and Constraints of Causative Serial Verb Constructions in Korean